Warning: Only Smart People Click Here?

so theoretically, i have multiple companies using my application, and each company has multiple users. Now data is stored in a sql database and certain parts of it are changed enough whereby it has to be "reselected" by every user after a specific amount of time. ill call this the "highly dynamic data." (I already sorted out the 'users by company' effeciently with a cluster index in the companyID column for the tables)

lets say that each session will "check" if it has to be updated every 15 seconds. How will it check? If someone submitted a change to the "highly dynamic data" in the sql database then they will also submit the "time of change" to an .xml database. Moreover, each session has a "last updated" session variable. The session that submitted the change will update his "last updated" session variable at the time of the sql update. However, all other sessions will have out dated "last updated" session variables. and every 15 seconds each session checks the xml database to see if their "last update" is current. If it is not current then it does the sql database update, else, it does nothing..is this a good mechanism for efficient updating? The goal is to keep the sql database as free as possible and reduce strain on the server?

edit: oh yeah, how would i implement the timing factor here? firstly, i assume the 15 second counter shouldnt be concurrent with all other users, otherwise the server would be a cluster-f*** every 15 seconds so how would i incorporate some randomness for each session?